fix(dn): a fragment of background paper no longer holds a page in grouping

Co-Authored-By: MiniMax Code <noreply@minimax.io>

src/lib/dn-pipeline/assemble.test.ts

@@ -51,6 +51,53 @@ describe("assemblePages", () => {
   });
 });
 
+describe("background fragments", () => {
+  const fragment = (over: Partial<PageRead> = {}) => page(1, {
+    multiple_documents: true,
+    items: [{ description: "Bend 90", qty: 7, unit: "ea", unit_price: null, line_amount: null, product_code: null }],
+    evidence: { multiple_documents: "A cropped corner of another paper is visible behind this invoice." },
+    ...over,
+  });
+
+  it("groups a complete document that only has background paper behind it", () => {
+    const result = assemblePages([fragment()]);
+    expect(result.groups.map((group) => group.pages)).toEqual([[1]]);
+    expect(result.openPages).toEqual([]);
+    expect(result.requiresReview).toBe(false);
+  });
+
+  it("still holds a page showing two complete documents side by side", () => {
+    const result = assemblePages([fragment({
+      evidence: { multiple_documents: "Two delivery notes side by side, each with its own header." },
+    })]);
+    expect(result.groups).toEqual([]);
+    expect(result.openPages).toEqual([{ page: 1, reason: "multiple_documents" }]);
+    expect(result.requiresReview).toBe(true);
+  });
+
+  it("still holds a background fragment on a page with no line items", () => {
+    const result = assemblePages([fragment({ items: [] })]);
+    expect(result.openPages).toEqual([{ page: 1, reason: "multiple_documents" }]);
+  });
+
+  it("still holds a background fragment on a page with no document number", () => {
+    const result = assemblePages([fragment({ doc_number: "" })]);
+    expect(result.openPages).toEqual([{ page: 1, reason: "multiple_documents" }]);
+  });
+
+  it("still holds the page when the evidence names a second document number", () => {
+    const result = assemblePages([fragment({
+      evidence: { multiple_documents: "Corner of invoice 22433700 visible behind, a second document." },
+    })]);
+    expect(result.openPages).toEqual([{ page: 1, reason: "multiple_documents" }]);
+  });
+
+  it("groups a fragment with no evidence text at all on a complete page", () => {
+    const result = assemblePages([fragment({ evidence: {} })]);
+    expect(result.openPages).toEqual([{ page: 1, reason: "multiple_documents" }]);
+  });
+});
+
 
 it("preserves manual split/merge decisions instead of rejoining by printed numbers",()=>{
  const reads=[page(1,{doc_number:"DN"}),page(2,{doc_number:"DN"}),page(3,{doc_number:"OTHER"})];

src/lib/dn-pipeline/assemble.ts

@@ -43,8 +43,10 @@ export function assemblePages(
   const readsByPage = new Map(ordered.map((read) => [read.page, read]));
   for (const read of ordered) {
     if (read.multiple_documents) {
-      openPages.push({ page: read.page, reason: "multiple_documents" });
-      continue;
+      if (!backgroundFragment(read)) {
+        openPages.push({ page: read.page, reason: "multiple_documents" });
+        continue;
+      }
     }
 
     const previous = groups[groups.length - 1];
@@ -113,6 +115,28 @@ export function assemblePages(
   return { groups, openPages, requiresReview: openPages.length > 0 };
 }
 
+const BACKGROUND = /background|behind|fragment|cropped|partial|edge|corner|stack/i;
+const DOCUMENT_NUMBER = /\d{4,}/g;
+/**
+ * True when a page flagged `multiple_documents` is really one complete
+ * document with a scrap of other paper behind it.
+ *
+ * The extraction prompt already tells the reader to ignore fragments, and it
+ * still flags them often enough to hold whole invoices in "Check grouping"
+ * for the want of a corner. Every condition has to hold: a number of its own
+ * and at least one line, so there is a document to group, and evidence that
+ * describes the other paper as background without naming a second reference —
+ * a second reference means there is a second document to decide about.
+ */
+function backgroundFragment(read: PageRead): boolean {
+  if (!read.doc_number.trim() || read.items.length === 0) return false;
+  const evidence = read.evidence?.multiple_documents;
+  if (!evidence || !BACKGROUND.test(evidence)) return false;
+  const own = [read.doc_number, ...(read.doc_number_alts ?? [])].join("|");
+  const elsewhere = evidence.replace(own, " ").match(DOCUMENT_NUMBER);
+  return elsewhere === null;
+}
+
 function sessionFor(page: number, hints: AssemblyHints): string | undefined {
   return hints.page_delivery_session_ids?.[page] ?? hints.delivery_session_ids?.[page] ??
     (hints.delivery_session_id ?? hints.deliverySessionId);